How ChatGPT and Google's AI choose which business to recommend
We asked three AI engines to recommend a business, and tracked where they got their answers. Here is what decides who gets named.
What we did
We took a common, high-intent question a real buyer would ask, "the best private chef in Nashville," and put it to the three engines people actually use to get recommendations today: ChatGPT (search mode), Perplexity, and Google's AI Overviews. For each, we recorded not just the answer, but the sources the engine cited to back that answer. We ran it across several phrasings of the same intent (best, hire, dinner party, small event) to see the pattern, not a single snapshot.
This is one category in one city. The point is not the specific businesses named, it is the shape of where the answers come from, and that shape is remarkably consistent across engines.
What the engines actually cited
Grouped by type, here is where the citations came from across the three engines:
| Source type | Examples cited | Share of citations |
|---|---|---|
| Best-of lists & roundups | Modern Luxury, local "best private chefs" guides | high |
| Booking directories | Take a Chef, Cozymeal, Chef Maison, PlatesFull | high |
| Review & marketplace platforms | Yelp, Thumbtack | medium |
| Social profiles | low | |
| The business's own website | a business's homepage | rare |
Why your website barely shows up
It feels backwards, so it is worth being precise about the mechanism. An AI assistant answering "who is the best X near me" is not ranking websites the way Google's blue links do. It is looking for consensus from sources it trusts, and a business talking about itself on its own site is the weakest possible signal of that. The engine wants to see that other people, on independent pages, name you. So the pages that get cited are the ones that list and compare many businesses: the roundups, the directories, the review sites.
That is why a beautiful website, on its own, does almost nothing for AI recommendations. The website matters for turning a visitor into a customer once they arrive. It is not what gets you named in the answer.
What this means if you want to be recommended
The takeaway is not "give up on your site." It is "put your effort where the citations actually are." Concretely:
- Get onto the exact pages the engines cite for your category. Not generic directories, the specific best-of lists and platforms that show up when you ask the AI your buyers' questions.
- Earn mentions on independent pages. Being named alongside recognized names, in content you do not control, is the strongest off-site signal there is.
- Keep your listings and reviews consistent across the directories and marketplaces, because that is the raw material the engines read.
- Measure it. The only way to know if it is working is to ask the engines your buyers' questions on a schedule and track how often you get named. That number is your share of voice.
None of this is fast, and none of it is a trick. It is the off-site authority that AI trusts, built deliberately. But right now, in most categories, almost no one is doing it, which is exactly why the window is open.
Common questions
Does my website matter at all for AI search?
Yes, but for a different job than most people think. Your site needs to be crawlable, clearly state who you are and where you operate, and convert the visitors who arrive. It is rarely the page an AI cites when recommending you. The citation comes from third-party sources.
How do I find which pages the AI cites for my business?
Ask the engines the real questions your customers would ask, and look at the sources they list under the answer. Those cited pages are your target list. Our free audit does this starting point for you.
Isn't this just SEO?
It overlaps, but the goal is different. Classic SEO earns a ranking in Google's links. This earns a mention inside the AI-generated answer that increasingly sits above them. The cited surface, and the work to get there, is more off-site than traditional SEO.
Want to see where you stand?
We will measure your current share of voice and show you which sources are getting cited instead of you. Free, no obligation.
Get my free auditMethod note: an informal probe of three AI answer engines (ChatGPT search, Perplexity, Google AI Overviews) for one local service category across several query phrasings, September 2026. AI answers are non-deterministic and vary by location and over time; this describes a consistent pattern we observed, not a precise ranking. Directional, not a controlled study.